Exposures to per- and polyfluoroalkyl substances in environmentally overburdened communities


A. Hernandez



April 11, 2023

Background


  • PFAS are ubiquitous in the environment, but the scope of contamination in the US is still broadly unknown

Motivating question


What are the characteristics of communities that are more exposed to PFAS through drinking water? Are exposures to PFAS in drinking water greater in environmental justice communities?

Background cont.


  • UCMR3 was the first and only national assessment of PFAS in PWSs to date

    • Testing was limited in scope (5 PFAS), high detection limits, not all PWSs represented
    • UCMR 5 testing will be more expansive, not completed until 2025
    • New EPA NPDWR will require more regular testing for 6 PFAS compounds
  • Some states have more expansive testing programs, but this work is incomplete, inconsistent, resource intensive.

Background cont.


  • Matching limited PWS monitoring data to demographic information presents a number of challenges, largely due to the lack of availability of information on service area boundaries at a national scale.

    • Spatial matching happens at the county level, but PWSs rarely serve an entire county.

Background cont.


How else might we assess exposure to PFAS?

  • Proximity to PFAS contamination sites is associated with detections of PFAS in drinking water (Hu et al. 2016)
  • Consumption of PFAS contaminated drinking water is associated with higher blood levels of PFAS (Post et al. 2021).


Background cont.


What is considered a “PFAS contamination site”? What facilities are using PFAS?

  • In the absence of comprehensive testing data, communities, researchers, and policymakers have turned to conceptual approaches which precautionarily consider certain types of facilities as “presumptive” or “potential” PFAS point sources

    • Presumptive facilities include wastewater treatment plants, current and former military sites, airports, and certain industrial facilities.

Study design/goals


Goal:

  • Assess the relationship between environmentally overburdened communities and PFAS point sources + contamination sites


Design:

  • Aggregate information about PFAS point sources and contamination sites
  • Match sites to a census tract
  • Assess socio-demographic and environmental characteristics at the census tract

PFAS point sources + contamination sites

PFAS point sources + contamination sites

Measuring/assessing environmental justice


EJ research on exposures to contaminants in drinking water has primarily relied on county-level demographics or has been limited to state-specific assessments for states with more comprehensive water system data.


This work has historically relied on unidimensional measures of socioeconomic status and race to evaluate disparities in environmental exposures.

Measuring/assessing environmental justice cont.


  • The Environmental Justice Index (EJI), developed by CDC and ATSDR, is a multi-dimensional index that accounts for cumulative exposures and disproportionate health outcomes

  • Compares census-tract level environmental, social, and health-based community indicators

  • Unlike EPA’s EJSCREEN and CDC’s Social Vulnerability Index, the EJI quantifies the cumulative impact of environmental burdens on health

  • Created to help identify communities experiencing disproportionate health outcomes associated with environmental exposures

Environmental Justice Index

Environmental Justice Index


  • The EJI underrepresents the health burden from exposures to contaminants (broadly, including PFAS) in drinking water and exposures to PFAS from other sources

    • This is mostly due to spatial mismatches/data incompleteness, but also because PFAS testing is limited and PFAS are generally unregulated/underreported
  • EJI does incorporate info about TRI reporting facilities (proportion of a census tract that falls within a 1-mile buffer of a TRI facility).

    • PFAS were added to the TRI program in 2020, however, only 39 facilities reported PFAS emissions in 2020, which is likely a substantial underestimation (Post et al. 2021, Reade et al. 2022).

Environmental Justice Index: Environmental Burden Module


Calculating the Env Burden:

  • Order the census tracts on a scale from 0 to 100 for a given variable to create a percentile rank for the percentage of tracts equal to or lower than the burden for that variable (0-1 index).

    • For variables like TRI sites, the metric is a proximity measure: the percent of the tract within 1-mile of a TRI site
  • An overall rank for each module is created by summing the individual variables within the module and repeating the percentile ranking process.

  • This is repeated again with the other two modules to calculate an EJI Rank

Methods

PFAS sources

Characteristic N = 35,5381
Dataset
    Part 139 Airports 532 (1.5%)
    EPA Stewardship Program Participating Facilities 16 (<0.1%)
    Facilities in Industries that may be Handling PFAS 31,902 (90%)
    Federal Agency Locations with Known or Suspected PFAS Detections 694 (2.0%)
    Fire Training Sites 448 (1.3%)
    Facilities that Manufacture or Import PFAS 807 (2.3%)
    Superfund Sites with PFAS Detections 894 (2.5%)
    Known PFAS Spills/Release Incidents 245 (0.7%)
1 n (%)

PFAS burden estimation

PFAS burden estimation: Adding a buffer

PFAS burden estimation: Combine the buffer

PFAS burden estimation: Find the intersection with a census tracts

PFAS burden estimation: Calculate % of census tract that intersects with PFAS buffer

\[ \text{perc_area_pfas}_i = \dfrac{\sum\text{area}(\text{PFAS point source with 1 mile buffer that overlaps with census tract i})}{\text{area}(\text{census tract i})}*100\% \]

PFAS burden estimation: % of census tract that intersects with PFAS buffer

PFAS burden estimation: % of census tract that intersects with PFAS buffer

PFAS burden estimation: Percentile rank of % of census tract within 1-mile buffer of PFAS source

PFAS burden map

PFAS burden estimation: Distribution

Table 2

Characteristic All Census Tracts Number of PFAS point sources
N = 71,680 0, N = 26,359 ≥ 1, N = 45,321 p-value1
% of census tract within 1-mile of a PFAS source
    Median (IQR) 6 (0 - 64)
    Mean (SD) 31 (39)
Sum of the HVM, EBM, and SVM module percentile ranks <0.001
    Median (IQR) 1.26 (0.82 - 1.80) 1.06 (0.66 - 1.58) 1.37 (0.93 - 1.93)
    Mean (SD) 1.33 (0.65) 1.14 (0.60) 1.43 (0.64)
Sum of the EBM variable percentile ranks <0.001
    Median (IQR) 6.09 (4.98 - 7.26) 5.20 (4.32 - 6.16) 6.64 (5.59 - 7.71)
    Mean (SD) 6.15 (1.61) 5.27 (1.35) 6.65 (1.53)
Sum of the SVM variable percentile ranks <0.001
    Median (IQR) 6.84 (5.24 - 8.42) 6.56 (5.02 - 8.03) 7.02 (5.39 - 8.62)
    Mean (SD) 6.81 (2.07) 6.53 (2.02) 6.97 (2.08)
Total number of HVM tertile flags <0.001
    Median (IQR) 1.00 (0.00 - 3.00) 1.00 (0.00 - 3.00) 1.00 (0.00 - 3.00)
    Mean (SD) 1.63 (1.54) 1.67 (1.55) 1.60 (1.53)
1 Wilcoxon rank sum test

Distribution of % of census tract within 1-mile of a PFAS source by quartiles of EJI modules

Correlation between PFAS sources and EBM variables

Correlation between PFAS sources and SVM variables

Correlation between PFAS sources and HVM variables

Table 3: Univariate logistic regression


Crude logistic regression models evaluating associations between presence of a PFAS point source/contamination site (Y/N) and EJI modules.

Characteristic N OR1 95% CI1 p-value
Sum of the EBM variable percentile ranks 71,680 1.92 1.90, 1.94 <0.001
Sum of the SVM variable percentile ranks 71,680 1.11 1.10, 1.12 <0.001
Total number of HVM tertile flags 71,680 0.97 0.96, 0.98 <0.001
1 OR = Odds Ratio, CI = Confidence Interval

Table 4: Multivariate logistic regression

Adjusted logistic regression models evaluating associations between presence of a PFAS point source/contamination site (Y/N) and EJI modules.

Characteristic OR1 95% CI1 p-value
Sum of the EBM variable percentile ranks 1.91 1.89, 1.93 <0.001
Sum of the SVM variable percentile ranks 1.10 1.09, 1.11 <0.001
Total number of HVM tertile flags 0.84 0.83, 0.85 <0.001
1 OR = Odds Ratio, CI = Confidence Interval

Recalculating the EJI


What’s the additional burden of PFAS to communities?

Additional burden of PFAS